Next POI recommendation is critical for personalized location-based services, yet existing methods struggle to jointly capture global structural patterns, sequential mobility dynamics, and semantic preference evolution. To address this problem, we propose HyBERT-Next-Next, a hypergraph-enhanced, temporally aware, multi-level sequential recommendation framework for next-POI prediction. HyBERT-Next-Next first constructs three adaptive POI hypergraphs from user check-in sequences and POI attributes, capturing high-order collaborative relations among locations. An adaptive Weisfeiler–Lehman refinement strategy is introduced to reduce structural redundancy of the hypergraphs while preserving higher-order connectivity, after which a hypergraph convolutional network learns global POI embeddings shared across users. To jointly capture long-range contextual dependencies and short-term transition continuity, an extended lightweight hybrid sequential encoder (eLHSE) combining both lightweight bidirectional Transformer (Light-BERT) and a bidirectional gated recurrent unit (BiGRU) is employed. An adaptive gating mechanism is applied to dynamically fuse the two representations at each time step, enabling the model to balance global structural context and local sequential regularities. The eLHSE is shared by three hierarchical channels corresponding to the venue, zone, and attribute levels, each followed by attention pooling to identify the most informative trajectory elements. The resulting multi-level user representations are fused through a gated mechanism with global hypergraph-based POI embeddings, producing a unified user–POI compatibility representation. HyBERT-Next is trained under a multi-task learning framework that simultaneously predicts the next POI, next zone, and next semantic attribute, allowing auxiliary tasks to regularize the backbone representations. Experimental results show that HyBERT-Next-Next outperforms state-of-the-art methods in accurately predicting the next POI, while also maintaining scalability and interpretability.

Vehicle platoon systems (VPSs) constitute a fundamental component of intelligent transportation systems (ITSs), enhancing traffic efficiency, safety, and energy conservation. This article presents an adaptive control strategy for unknown nonlinear heterogeneous VPSs under a bidirectional (BD) communication topology. The proposed approach simultaneously addresses asymmetric spacing constraints, actuator saturation, and dead-zone nonlinearities. First, an equivalent transformation is adopted to eliminate the need for precise modeling of both system dynamics and actuator nonlinearities. An adaptive mechanism is then designed to jointly estimate the bounds of approximation errors, the norm of ideal NN weights, and the derivative bounds of control inputs. Based on both the estimated bounds and the gap about the time-headway policy, a novel controller is constructed to enhance safety and vehicle stability. Furthermore, a barrier Lyapunov function (BLF) is incorporated to rigorously enforce intervehicle spacing constraints, thereby ensuring collision avoidance. The proposed control scheme is theoretically proven and numerically validated to guarantee both individual vehicle stability and string stability of the platoon.

In this paper, the problem of proportional-integral observer (PIO) design is investigated for a class of discrete-time multi-rate systems with multiple sensors, with the sensor sampling periods being allowed to differ from the system updating periods. The facilitation of communication between sensors and the remote PIO through wireless networks, which are subject to probabilistic packet dropouts, is achieved through the utilization of a decode-and-forward relay-based strategy. The occurrence of packet dropouts is governed by a Bernoulli-distributed random variable whose probability is dependent on the available transmission power. A decode-and-forward relay-based strategy, developed based on different components, is capable of processing information from different encoders at different physical locations. For the convenience of observer design, the lifting technique is employed with aim to cast the multi-rate system into a single-rate one. By establishing sufficient conditions, the combined effect of external noises and relaying-aided communication on estimation performance is intuitively illustrated. Subsequently, a PIO with an adjustable parameter is designed by solving certain optimization problems. A simulation example is finally provided to validate the theoretical results.

The federated-filtering-based (FFB) fusion estimation problem is investigated in this paper for networked multi-rate systems, where the measurement signals are transmitted over a wireless network with limited transmission power. A probabilistic quantization mechanism is introduced to handle the raw measurement signals for the purpose of facilitating digital communication over network. Certain transmission models are proposed to describe the behaviors under the effects of multi-rate dynamics, probabilistic quantization and limited transmission power. A delicately designed FFB fusion scheme is proposed to acquire the desired state estimates, where the local filters will receive feedback from the fusion center to reset their estimates. The parameters for the local filters are calculated by recursively minimizing their upper-bounds for the estimation error covariances. Furthermore, new conditions have been derived to analyze the ultimately boundedness of the estimation error covariance for the fusion center. Subsequently, a power allocation strategy is designed by minimizing such ultimate bound subject to the given transmission power constraint. Finally, the effectiveness of the proposed fusion estimation strategy and its optimal power allocation scheme is verified through a simulation example. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Microgrids (MGs) are continuously evolving with the advancement of distributed energy storage devices and renewable energy generation. However, limited bandwidth and denial-of-service (DoS) attacks are threatening the stable operation of MGs by impacting data availability. By characterizing DoS attacks with attack frequency and duration, this paper pays attention to the secondary voltage restoration issue of a discrete MG by utilizing proportional-integral observers (PIOs) and dynamic encoding-decoding mechanisms (DEDMs). First, a local PIO with a predefined forgetting coefficient in its accumulative term is employed to estimate the states of the discrete-time MG, thereby mitigating the impact of DoS attacks. Subsequently, a consensus-based secondary controller is developed for the addressed system, leveraging the feedback linearization and PIO. DEDMs are implemented in communication channels to ensure efficient and reliable data exchanges. By resorting to the Lyapunov stability theory, sufficient conditions for voltage restoration are derived, taking into account the average dwell time and the proportion of attack silence. Moreover, desired observer and controller gains are calculated by solving a couple of matrix inequalities. Finally, a series of simulations based on Simulink are conducted to verify the practicality of the proposed controller.

In this paper, a neural network (NN)-based joint state-disturbance observer is investigated for uncrewed surface vehicles (USVs) deployed an encoding-decoding mechanisms (EDMs). To reduce the communication burden, an EDM with recoverable encoded data is employed on the sensor-observer transmission channel, transforming measurement signals into binary code formats. Under this schedule, a joint state-disturbance observer is developed via NNs to estimate both the USV’s nonlinearities and the hybrid disturbances due to parameter perturbations and environmental uncertainties in marine operations. The NN weight update law associated with the EDM is developed with the help of gradient descent. Furthermore, sufficient conditions on the NN learning rates and auxiliary coefficients are derived via Lyapunov theory to guarantee the uniformly ultimately bounded (UUB) convergence of both the NN weights and estimation errors. And the gain of NN-based observer under EDMs is calculated by resolving matrix inequality. Finally, the Cybership II model vessel is used to illustrate the effectiveness of the proposed joint observer.

This paper is concerned with the platooning control problem of connected automated vehicles (CAVs) under non-uniform stochastic vehicle-to-vehicle (V2V) communication delays. Most existing relevant studies assume uniform or deterministic or slowly varying delays, or design platoon controllers based on worst-case delay bounds, resulting in overly conservative analysis and design criteria. To address this gap, we first develop a stochastic delay model that characterizes heterogeneous and time-varying delays across multiple V2V links using statistical distributions. Building on this model, we then propose a distributed platooning control strategy that explicitly incorporates the stochastic delay characteristics and a refined constant-time headway spacing policy into the control law, enabling robust tracking performance while ensuring both individual stability and string stability. Furthermore, rigorous stability analysis allows us to establish sufficient delay-distribution-dependent conditions linking delay statistics to controller gain design, providing insight into the trade-offs between robustness and performance. Finally, extensive simulation studies are provided to demonstrate the efficacy of the proposed platooning control method.

Monocular video-based 3D face tracking is vital for interactive pattern recognition and human avatars. Most existing image-based methods fail to model temporal dependencies in video, causing jitter and inaccuracies. Furthermore, they also often neglect the continuous multi-modal signals present in facial videos such as expression dynamics and emotional cues that provide essential temporal drivers for facial modeling. To this end, this study first explores the Mamba architecture tailored for 3D facial tracking by proposing a hierarchical Mamba framework, termed HMamba-3DFT. The proposed network can efficiently capture and track variations in 3D facial shapes from a monocular video. To exploit the global spatiotemporal correlations across frames of the dynamic face, we develop a bidirectional spatiotemporal vision Mamba (BSTV-Mamba) module featuring a bidirectional spatiotemporal selective scan (BSTS-Scan) mechanism. To capture temporally evolving multi-modal emotion signals embedded in continuous video sequences, we introduce a dynamic emotion-driven mechanism. Additionally, to mitigate the potential degradation of reconstruction fidelity caused by an over-reliance on emotion-driven cues, we integrate facial semantic alignment with facial emotion driving to enhance the accuracy of emotion-driven facial modeling. This integrated dual-optimization strategy systematically guides the network during training, ensuring that the reconstructed 3D facial mesh not only accurately captures the emotional attributes of the input frames but also benefits from enhanced optimization for more precise reconstruction. Extensive evaluations on benchmark datasets show competitive performance against state-of-the-art methods.

This article investigates the probability-guaranteed distributed fusion estimation (DFE) issue for a class of nonlinear systems with censored measurements over sensor networks. In order to reduce the transmission load and minimize energy consumption, each sensing node sends its censored measurement to the estimator only when a predefined event-triggered media access condition is met. To enhance both the security and reliability of signal transmission, a binary encoding scheme is adopted, which encodes the transmitted information using a limited number of bits. The core objective of the addressed problem is to develop a novel distributed robust fusion estimator capable of maintaining effectiveness in the presence of censored measurements, nonlinear dynamics, and unknown-but-bounded noise, while confining the estimation error to ellipsoidal regions with a prespecified probability. To be specific, sufficient conditions for the existence of the desired fusion estimator, obtained through a weighted matrix fusion technique, are established using mathematical induction, linear matrix inequalities, and set theory. In addition, the estimator parameters and the fusion weights are determined through the solution of a series of optimization problems that are constrained by matrix inequalities. Finally, the efficacy of the DFE scheme developed in the article is confirmed by a ballistic object tracking example.

This article addresses the problem of input-output data-based ultimate boundedness control for a class of networked systems subject to probabilistic bit flips and false data injection (FDI) attacks under the try-once-discard (TOD) protocol. First, a prior experiment is conducted to obtain a set of input-output data from the considered system, which will be utilized for the data-based controller design. A uniform-quantization-based encoding-decoding mechanism is employed to digitalize measurement signals. The TOD protocol is adopted to schedule signal transmissions between encoders and decoders. Considering the nature of digital communication, an ellipsoid constraint and a sequence of Bernoulli variables are introduced to account for the FDI attacks and bit flips during the transmission, respectively. To expediently design the data-based controller, a novel autoregression (AR)-based method is proposed, subject to probabilistic bit flips and protocol-induced effects. This article aims to design a data-driven controller that ensures the ultimate boundedness of the closed-loop system under the effects of TOD protocol scheduling and communication failures. Sufficient conditions are presented to ensure the desired control performance by using the S-Lemma from data. An improved cone complementarity linearization (CCL) algorithm is developed to calculate the controller gain. Eventually, a numerical simulation example is provided to demonstrate the effectiveness and feasibility of the proposed data-based ultimate boundedness control scheme.

Frequency stability is vital for the reliable operation of critical equipment in power systems. However, increasing renewable integration and unpredictable actuator faults hinder traditional controllers from achieving frequency regulation within a fixed and predetermined time. To overcome such a challenge, a novel resilient decentralized fixed-time dynamic feedback controller is proposed for multiarea power systems subject to nonhomogeneous Markovian jumps, actuator faults, and load fluctuations. In contrast to existing fixed-time control approaches dependent on state change rates, the proposed dynamic controller is uniquely governed by both control amplitude bounds and quadratic state terms. The designed controller rigorously guarantees both the stochastic fixed-time stability and the L-infinity performance despite simultaneous actuator failures and nonhomogeneous parameter jumps. By applying Lyapunov differential inequalities and set measure theory, a tractable design criterion is established to facilitate the solution of desired gain matrices, ensuring scalability and plug-and-play functionality. A three-area power system is finally utilized to evaluate the effectiveness of the developed control strategy regarding fixed-time convergence and resilience against load fluctuations and actuator faults.

ABSTRACT This paper investigates the issue of set‐membership filtering (SMF) for time‐varying multirate systems with censored measurements. To meet practical transmission requirements, a multirate strategy is proposed to regulate the sampling rates of controlled devices and sensors. Furthermore, an encoding–decoding mechanism (EDM) is introduced to address the potential constraints on communication bandwidth and energy resources in industrial systems. By using the lifting technique, a filtering error is established based on a uniform sampling rate. With recursive linear matrix inequalities (RLMI), a sufficient condition for the realizability of the SMF is derived under the multirate mechanism, bounded noises, censored measurements as well as EDMs. Such a condition determines an ellipsoidal region that contains all possible real states at each time step. In addition, an optimization model utilizing the matrix trace as the metric is formulated under the corresponding inequality constraints, thereby yielding the ellipsoidal region of minimal volume. Finally, numerical simulations are used to verify that our designed filter is effective.

Deep learning-based visual trackers have made remarkable progress and achieved outstanding performance. Although existing lightweight trackers typically improve efficiency by adopting compact CNN architectures, neural architecture search, pruning, low-resolution processing, or simplified correlation operations, these approaches may limit their ability to model long-range dependencies between the template and search regions. Similarly, although bidirectional Transformer mechanisms enable comprehensive information exchange between the two branches, they introduce unnecessary computational overhead for lightweight tracking and increase background interference from the search region. To address these limitations, we propose a lightweight Siamese-like Transformer tracker, named UMATracker, based on split-head Unidirectional Mixed Attention. Specifically, a lightweight Siamese-like feature extraction backbone based on transformers is proposed to extract features efficiently in parallel. Then, an efficient multi-head mixed module embedded with some split-head unidirectional mixed attention layers is created to perform the association and integration between template and search region features, enhancing the tracking target information transmission from the template branch to the search branch. In this way, the proposed module efficiently transfers target information from the template branch to the search branch while reducing redundant interactions. Furthermore, we introduce a distribution-level knowledge distillation strategy for lightweight tracking that transfers the probability distributions of the top-left and bottom-right corners predicted by a stronger teacher tracker. This soft supervision helps the lightweight model learn more useful localization information, thereby improving tracking accuracy and robustness. Finally, several lightweight and general trackers are used to conduct comparative experiments on multiple tracking benchmarks. Results clearly verify that our tracker can achieve superior performance with fewer parameters and a faster running speed.

Microgrids (MGs) have emerged in response to the rapid development of renewable energy generation, in which distributed secondary control is of great importance to guarantee the power quality under the hierarchical control framework. Cyberattacks and unknown inputs present significant challenges, primarily due to network vulnerabilities and environmental complexity. Considering typical replay attacks (RAs) with certain attack frequency and duration, this paper concentrates on realizing voltage restoration for MGs with the assistance of the unknown input observer (UIO) and triggered encoding-decoding mechanisms (TEDMs). First, a TEDM-based UIO is proposed to simultaneously estimate the system states and unknown inputs. Subsequently, a distributed secondary voltage controller with unknown input compensation is designed based on the locally estimated unknown inputs and the estimated states of neighboring systems affected by RAs. Moreover, in light of the average dwell time and the silence proportion of RAs, sufficient conditions are derived to ensure the bounded voltage restoring performance. Accordingly, the observer and controller gains can be obtained through several matrix inequalities. Finally, several test cases are implemented in MATLAB/Simulink to validate the performance of the proposed approach.

The distributed energy management (DEM) problem of smart grids is devoted to achieving optimal energy dispatch and allocation to ensure social welfare maximization. It can be modelled as a distributed optimization problem with physical constraints, whose solution depends on data sharing between smart devices. The exchange of information creates a potential risk for eavesdroppers to intercept and access private data. To avoid privacy leakages, a privacy-preserving optimization strategy via output masks in a consensus framework is proposed to realize social welfare maximization of energy management, where power demand relies on the parameters of the demand sides. Theoretical analysis discloses that eavesdroppers are unable to accurately infer private information of the generators/demand loads. Furthermore, both the convergence and optimality of the proposed strategy are discussed by means of the matrix perturbation theory and the famous KKT optimality condition. Finally, the effectiveness of the proposed strategy for solving the DEM problem is confirmed by the simulation experiments.

Visual object tracking (VOT) plays a pivotal role in unmanned aerial vehicle (UAV) applications. Addressing the trade-off between accuracy and efficiency, especially under challenging conditions like unpredictable occlusion, remains a significant challenge. This paper introduces LGTrack, a unified UAV tracking framework that integrates dynamic layer selection, efficient feature enhancement, and robust representation learning for occlusions. By employing a novel lightweight Global-Grouped Coordinate Attention (GGCA) module, LGTrack captures long-range dependencies and global contexts, enhancing feature discriminability with minimal computational overhead. Additionally, a lightweight Similarity-Guided Layer Adaptation (SGLA) module replaces knowledge distillation, achieving an optimal balance between tracking precision and inference efficiency. Experiments on three datasets demonstrate LGTrack's state-of-the-art real-time speed (258.7 FPS on UAVDT) while maintaining competitive tracking accuracy (82.8% precision). Code is available at https://github.com/XiaoMoc/LGTrack

The rapid development of intelligent connected vehicles has led to widespread attention for platooning control, which is an effective method to mitigate a variety of societal issues. This article is concerned with platooning control via a refined constant-time-headway (CTH) strategy in the framework of model predictive control (MPC), where round-robin (RR) protocols are introduced to alleviate the communication burden. First, a dynamic model for platoon tracking errors is developed by using matrix transformation to account for the influence of the refined CTH strategy and the RR protocol. With the help of stability analysis, the original optimization problem involving unknown disturbances is transformed into an auxiliary MPC optimization problem to minimize its upper bound. Considering the cyclic characteristics of RR protocols, some sufficient conditions are then acquired to ensure the recursive feasibility of the MPC optimization problem. The desired controller parameters are obtained by means of an online optimization algorithm. Furthermore, the stability of platooning systems is disclosed under the developed sufficient conditions. Finally, the effectiveness of the devised control scheme is evaluated through numerical simulations.

Platooning control, an important technology in intelligent transportation systems, demonstrates its superior ability to improve traffic efficiency. This paper focuses on the issue of collision-free platooning control for automated vehicles, where an encryption-decryption-based protocol is employed to ensure communication security. First, a proportional integral observer (PIO) with memory fading cumulative sum about the estimated output errors is constructed to estimate vehicle dynamic information. A collision-free control scheme is developed by using an auxiliary control signal within the framework of artificial potential fields according to the estimated dynamics. A sufficient condition is achieved for the safe platoon requirement with the help of Lyapunov stability theory. Furthermore, the desired gains of both the collision-free controller and the PIO are determined by two particular matrix inequalities, which are solely connected to the maximum and minimum eigenvalues of associated communication topologies. To validate the effectiveness of the safety control scheme, a detailed simulation experiment is conducted on the Carsim/Simulink co-simulation platform.